I only follow some YouTubers like Digital Spaceport but there has been a lot of progress from years ago when LLM’s where only predictive. They now have an inductive engine attached to the LLM to provide logic guard rails.